open-thought/reasoning-gym
1
0
Fork 0
mirror of https://github.com/open-thought/reasoning-gym.git synced 2026-04-26 17:13:17 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

change parameter order for basic arc tasks

Browse source
This commit is contained in:
Andreas Koepf 2025-02-02 17:25:37 +01:00
parent 2409b3cda2
commit 604db012c3
2 changed files with 292 additions and 239 deletions
Download patch file Download diff file Expand all files Collapse all files

61
tests/test_arc_1d.py
View file

@ -1,36 +1,59 @@
import random
import pytest
from reasoning_gym.cognition.arc_1d import (
task_move_n_pix, task_move_n_pix_wrapped, task_gravity, task_gravity_counting,
task_gravity_antigravity, task_block_touch_dot, task_block_touch_dot_n_pix,
task_block_scale_to_dot, task_two_points_and_fill, task_reflect_block_with_border_pixel,
task_reflect_block_with_border_pixel_random, task_reflect_block_around_dot,
task_block_and_noise_remove, task_block_and_noise_remove_inside,
task_copy_block_to_dots, task_copy_block_to_dots_colors, task_paint_biggest_block,
task_sort_blocks_by_size, task_sort_complete_sequence, task_recolor_blocks_by_size,
task_gravity_one_step, task_move_block_by_own_size, task_change_to_five,
task_recolor_blocks_from_palette, task_duplicate_block_from_seeds,
task_fill_from_pixel, task_mark_size_two_blocks, task_fill_until_collision,
task_repeat_pattern_full, task_gravity_weighted_colors, task_color_left_half_blocks
task_block_and_noise_remove,
task_block_and_noise_remove_inside,
task_block_scale_to_dot,
task_block_touch_dot,
task_block_touch_dot_n_pix,
task_change_to_five,
task_color_left_half_blocks,
task_copy_block_to_dots,
task_copy_block_to_dots_colors,
task_duplicate_block_from_seeds,
task_fill_from_pixel,
task_fill_until_collision,
task_gravity,
task_gravity_antigravity,
task_gravity_counting,
task_gravity_one_step,
task_gravity_weighted_colors,
task_mark_size_two_blocks,
task_move_block_by_own_size,
task_move_n_pix,
task_move_n_pix_wrapped,
task_paint_biggest_block,
task_recolor_blocks_by_size,
task_recolor_blocks_from_palette,
task_reflect_block_around_dot,
task_reflect_block_with_border_pixel,
task_reflect_block_with_border_pixel_random,
task_repeat_pattern_full,
task_sort_blocks_by_size,
task_sort_complete_sequence,
task_two_points_and_fill,
)
def test_all_arc_1d_tasks():
"""Test that all ARC 1D task functions can be executed without exceptions."""
rng = random.Random(42) # Fixed seed for reproducibility
size = 20 # Reasonable size for testing
# Test all task functions
# Fixed move_pix value for testing
move_pix = 2
tasks = [
(lambda s, r, **k: task_move_n_pix(s, move_pix, k["solid"], r), {"solid": True}),
(lambda s, r, **k: task_move_n_pix_wrapped(s, move_pix, k["solid"], r), {"solid": True}),
(task_move_n_pix, {"move_pix": move_pix, "solid": True}),
(task_move_n_pix_wrapped, {"move_pix": move_pix, "solid": True}),
(task_gravity, {}),
(task_gravity_counting, {}),
(task_gravity_antigravity, {}),
(task_block_touch_dot, {}),
(lambda s, r, **k: task_block_touch_dot_n_pix(s, move_pix, r), {}),
(task_block_touch_dot_n_pix, {"move_pix": move_pix}),
(task_block_scale_to_dot, {}),
(task_two_points_and_fill, {}),
(task_reflect_block_with_border_pixel, {}),
@ -54,7 +77,7 @@ def test_all_arc_1d_tasks():
(task_fill_until_collision, {}),
(task_repeat_pattern_full, {}),
(task_gravity_weighted_colors, {}),
(task_color_left_half_blocks, {})
(task_color_left_half_blocks, {}),
]
for task_func, kwargs in tasks:
@ -62,7 +85,7 @@ def test_all_arc_1d_tasks():
success = False
for _ in range(10): # Try up to 10 times
try:
result = task_func(size, rng, **kwargs)
result = task_func(rng, size, **kwargs)
if result is not None:
success = True
# Basic structure checks
@ -74,5 +97,5 @@ def test_all_arc_1d_tasks():
break
except Exception as e:
pytest.fail(f"Task {task_func.__name__} failed with error: {str(e)}")
assert success, f"Task {task_func.__name__} always returned None in 10 attempts"